Anthropic signed a $200 million Pentagon contract in July, but talks stalled when the DOD wanted unfettered access to Claude for all lawful purposes while Anthropic insisted its technology not be used for fully autonomous weapons or domestic mass surveillance.
The Pentagon told all its contractors they couldn't do any commercial business with the company. That includes Amazon and Google, who provide Anthropic's cloud compute. Strip that away and the company doesn't just lose government revenue. It loses the ability to operate. When the U.S. government wanted to do something similar to Huawei, Congress had to pass a law. So the court said no.
OpenAI signed a Pentagon deal with functionally similar safety restrictions hours after Anthropic was blacklisted for those same restrictions. The framing was different (cite existing laws vs. demand explicit contract terms) but the substance is close enough that it makes the "national security" rationale look weak.
I'd keep one eye on the Ninth Circuit appeal window (seven days) and the separate DC Circuit lawsuit over the civilian agency designation.